The final version

- more Scheduler
- New painting interface
- New any model inpaiting support
- Variable Resolution
- etc.

Mochi Diffusion.xcodeproj/project.pbxproj

@@ -16,7 +16,6 @@
 		03173C132999E2B500B03456 /* SDModel.swift in Sources */ = {isa = PBXBuildFile; fileRef = 03173C122999E2B500B03456 /* SDModel.swift */; };
 		03173C152999F5C700B03456 /* ImageController.swift in Sources */ = {isa = PBXBuildFile; fileRef = 03173C142999F5C700B03456 /* ImageController.swift */; };
 		03173C18299B3C1300B03456 /* ImageStore.swift in Sources */ = {isa = PBXBuildFile; fileRef = 03173C17299B3C1300B03456 /* ImageStore.swift */; };
-		0352E2A5294E2591003FBF25 /* Sparkle in Frameworks */ = {isa = PBXBuildFile; productRef = 0352E2A4294E2591003FBF25 /* Sparkle */; };
 		0352E2A7294E3148003FBF25 /* HelpCommands.swift in Sources */ = {isa = PBXBuildFile; fileRef = 0352E2A6294E3148003FBF25 /* HelpCommands.swift */; };
 		0352E2AA294EA0E1003FBF25 /* AppView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 0352E2A9294EA0E1003FBF25 /* AppView.swift */; };
 		0352E2AE294EA2B4003FBF25 /* MessageBanner.swift in Sources */ = {isa = PBXBuildFile; fileRef = 0352E2AD294EA2B4003FBF25 /* MessageBanner.swift */; };
@@ -61,6 +60,9 @@
 		C1ADEC2C2B16957800E142CA /* FolderMonitor.swift in Sources */ = {isa = PBXBuildFile; fileRef = C1ADEC2B2B16957800E142CA /* FolderMonitor.swift */; };
 		D7B03F2029D42F9900DF89DD /* SDModelAttentionType.swift in Sources */ = {isa = PBXBuildFile; fileRef = D7B03F1F29D42F9900DF89DD /* SDModelAttentionType.swift */; };
 		E45FA5822B7F7E4B009E90F0 /* GuernikaKit in Frameworks */ = {isa = PBXBuildFile; productRef = E45FA5812B7F7E4B009E90F0 /* GuernikaKit */; };
+		E4EC46322B890B6D00351E8C /* Sparkle in Frameworks */ = {isa = PBXBuildFile; productRef = E4EC46312B890B6D00351E8C /* Sparkle */; };
+		E4F15B1D2B86720F00E1EC3C /* de-coremldata.bin in Resources */ = {isa = PBXBuildFile; fileRef = E4F15B1C2B86720F00E1EC3C /* de-coremldata.bin */; };
+		E4F15B1F2B8673DB00E1EC3C /* en-coremldata.bin in Resources */ = {isa = PBXBuildFile; fileRef = E4F15B1E2B8673DB00E1EC3C /* en-coremldata.bin */; };
 /* End PBXBuildFile section */
 
 /* Begin PBXFileReference section */
@@ -132,6 +134,8 @@
 		C1220FBF2AFB122F007E5055 /* ImageWellView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ImageWellView.swift; sourceTree = "<group>"; };
 		C1ADEC2B2B16957800E142CA /* FolderMonitor.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = FolderMonitor.swift; sourceTree = "<group>"; };
 		D7B03F1F29D42F9900DF89DD /* SDModelAttentionType.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = SDModelAttentionType.swift; sourceTree = "<group>"; };
+		E4F15B1C2B86720F00E1EC3C /* de-coremldata.bin */ = {isa = PBXFileReference; lastKnownFileType = archive.macbinary; path = "de-coremldata.bin"; sourceTree = "<group>"; };
+		E4F15B1E2B8673DB00E1EC3C /* en-coremldata.bin */ = {isa = PBXFileReference; lastKnownFileType = archive.macbinary; path = "en-coremldata.bin"; sourceTree = "<group>"; };
 /* End PBXFileReference section */
 
 /* Begin PBXFrameworksBuildPhase section */
@@ -140,8 +144,8 @@
 			buildActionMask = 2147483647;
 			files = (
 				036BFC4E294B9FDB00D8AD04 /* Path in Frameworks */,
+				E4EC46322B890B6D00351E8C /* Sparkle in Frameworks */,
 				E45FA5822B7F7E4B009E90F0 /* GuernikaKit in Frameworks */,
-				0352E2A5294E2591003FBF25 /* Sparkle in Frameworks */,
 				0388DEB0297B00FC008B1C1C /* CompactSlider in Frameworks */,
 			);
 			runOnlyForDeploymentPostprocessing = 0;
@@ -275,6 +279,8 @@
 			isa = PBXGroup;
 			children = (
 				036BFC25294B9F7600D8AD04 /* Assets.xcassets */,
+				E4F15B1E2B8673DB00E1EC3C /* en-coremldata.bin */,
+				E4F15B1C2B86720F00E1EC3C /* de-coremldata.bin */,
 				03FD2318295F74B6006EEEE2 /* RealESRGAN.mlmodel */,
 			);
 			path = Resources;
@@ -298,9 +304,9 @@
 			name = "Mochi Diffusion";
 			packageProductDependencies = (
 				036BFC4D294B9FDB00D8AD04 /* Path */,
-				0352E2A4294E2591003FBF25 /* Sparkle */,
 				0388DEAF297B00FC008B1C1C /* CompactSlider */,
 				E45FA5812B7F7E4B009E90F0 /* GuernikaKit */,
+				E4EC46312B890B6D00351E8C /* Sparkle */,
 			);
 			productName = "Mochi Diffusion";
 			productReference = 036BFC1E294B9F7500D8AD04 /* Mochi Diffusion.app */;
@@ -367,9 +373,9 @@
 			mainGroup = 036BFC15294B9F7500D8AD04;
 			packageReferences = (
 				036BFC4C294B9FDB00D8AD04 /* XCRemoteSwiftPackageReference "Path.swift" */,
-				0352E2A3294E2591003FBF25 /* XCRemoteSwiftPackageReference "Sparkle" */,
 				0388DEAE297B00FC008B1C1C /* XCRemoteSwiftPackageReference "CompactSlider" */,
 				E45FA5802B7F7E4B009E90F0 /* XCRemoteSwiftPackageReference "GuernikaKit" */,
+				E4EC46302B890B6D00351E8C /* XCRemoteSwiftPackageReference "Sparkle" */,
 			);
 			productRefGroup = 036BFC1F294B9F7500D8AD04 /* Products */;
 			projectDirPath = "";
@@ -389,6 +395,8 @@
 				03F578642967CE9A003A815F /* Localizable.strings in Resources */,
 				0311C6BD2989E3EF0074BCAE /* Localizable.stringsdict in Resources */,
 				036BFC26294B9F7600D8AD04 /* Assets.xcassets in Resources */,
+				E4F15B1F2B8673DB00E1EC3C /* en-coremldata.bin in Resources */,
+				E4F15B1D2B86720F00E1EC3C /* de-coremldata.bin in Resources */,
 			);
 			runOnlyForDeploymentPostprocessing = 0;
 		};
@@ -700,14 +708,6 @@
 /* End XCConfigurationList section */
 
 /* Begin XCRemoteSwiftPackageReference section */
-		0352E2A3294E2591003FBF25 /* XCRemoteSwiftPackageReference "Sparkle" */ = {
-			isa = XCRemoteSwiftPackageReference;
-			repositoryURL = "https://github.com/sparkle-project/Sparkle";
-			requirement = {
-				kind = upToNextMajorVersion;
-				minimumVersion = 2.0.0;
-			};
-		};
 		036BFC4C294B9FDB00D8AD04 /* XCRemoteSwiftPackageReference "Path.swift" */ = {
 			isa = XCRemoteSwiftPackageReference;
 			repositoryURL = "https://github.com/mxcl/Path.swift.git";
@@ -732,14 +732,17 @@
 				minimumVersion = 1.5.0;
 			};
 		};
+		E4EC46302B890B6D00351E8C /* XCRemoteSwiftPackageReference "Sparkle" */ = {
+			isa = XCRemoteSwiftPackageReference;
+			repositoryURL = "https://github.com/sparkle-project/Sparkle.git";
+			requirement = {
+				kind = upToNextMajorVersion;
+				minimumVersion = 2.5.2;
+			};
+		};
 /* End XCRemoteSwiftPackageReference section */
 
 /* Begin XCSwiftPackageProductDependency section */
-		0352E2A4294E2591003FBF25 /* Sparkle */ = {
-			isa = XCSwiftPackageProductDependency;
-			package = 0352E2A3294E2591003FBF25 /* XCRemoteSwiftPackageReference "Sparkle" */;
-			productName = Sparkle;
-		};
 		036BFC4D294B9FDB00D8AD04 /* Path */ = {
 			isa = XCSwiftPackageProductDependency;
 			package = 036BFC4C294B9FDB00D8AD04 /* XCRemoteSwiftPackageReference "Path.swift" */;
@@ -755,6 +758,11 @@
 			package = E45FA5802B7F7E4B009E90F0 /* XCRemoteSwiftPackageReference "GuernikaKit" */;
 			productName = GuernikaKit;
 		};
+		E4EC46312B890B6D00351E8C /* Sparkle */ = {
+			isa = XCSwiftPackageProductDependency;
+			package = E4EC46302B890B6D00351E8C /* XCRemoteSwiftPackageReference "Sparkle" */;
+			productName = Sparkle;
+		};
 /* End XCSwiftPackageProductDependency section */
 	};
 	rootObject = 036BFC16294B9F7500D8AD04 /* Project object */;

Mochi Diffusion.xcodeproj/project.xcworkspace/xcshareddata/swiftpm/Package.resolved

@@ -14,8 +14,8 @@
       "kind" : "remoteSourceControl",
       "location" : "https://github.com/GuernikaCore/GuernikaKit.git",
       "state" : {
-        "revision" : "2c1ce26535279b309149acba47ae65e6a9394cb5",
-        "version" : "1.5.0"
+        "revision" : "dae44bf061ffa84df7f882cb563070d484488e44",
+        "version" : "1.6.0"
       }
     },
     {
@@ -41,14 +41,14 @@
       "kind" : "remoteSourceControl",
       "location" : "https://github.com/GuernikaCore/Schedulers.git",
       "state" : {
-        "revision" : "5e6ab9fa91aa56fc34edc6bae9de1caa7f13deb8",
-        "version" : "1.2.0"
+        "revision" : "1f517514d679e38bb9915c3a74bf04f75d5b5875",
+        "version" : "1.4.0"
       }
     },
     {
       "identity" : "sparkle",
       "kind" : "remoteSourceControl",
-      "location" : "https://github.com/sparkle-project/Sparkle",
+      "location" : "https://github.com/sparkle-project/Sparkle.git",
       "state" : {
         "revision" : "47d3d90aee3c52b6f61d04ceae426e607df62347",
         "version" : "2.5.2"

Mochi Diffusion/Model/SDModel.swift

@@ -17,7 +17,7 @@ struct SDModel: Identifiable {
     let attention: SDModelAttentionType
     let controlNet: [String]
     let isXL: Bool
-    let inputSize: CGSize?
+    var inputSize: CGSize?
 
     var id: URL { url }
 
@@ -103,8 +103,7 @@ private func identifyIfXL(_ url: URL) -> Bool {
 private func unetMetadataURL(from url: URL) -> URL? {
     let potentialMetadataURLs = [
         url.appending(components: "Unet.mlmodelc", "metadata.json"),
-        url.appending(components: "UnetChunk1.mlmodelc", "metadata.json"),
-        url.appending(components: "ControlledUnet.mlmodelc", "metadata.json")
+        url.appending(components: "UnetChunk1.mlmodelc", "metadata.json")
     ]
 
     return potentialMetadataURLs.first {
@@ -113,11 +112,11 @@ private func unetMetadataURL(from url: URL) -> URL? {
 }
 
 private func identifyInputSize(_ url: URL) -> CGSize? {
-    let encoderMetadataURL = url.appending(path: "VAEEncoder.mlmodelc").appending(path: "metadata.json")
+    let encoderMetadataURL = url.appending(path: "VAEDecoder.mlmodelc").appending(path: "metadata.json")
     if let jsonData = try? Data(contentsOf: encoderMetadataURL),
         let jsonArray = try? JSONSerialization.jsonObject(with: jsonData) as? [[String: Any]],
         let jsonItem = jsonArray.first,
-        let inputSchema = jsonItem["inputSchema"] as? [[String: Any]],
+        let inputSchema = jsonItem["outputSchema"] as? [[String: Any]],
         let controlnetCond = inputSchema.first,
         let shapeString = controlnetCond["shape"] as? String {
         let shapeIntArray = shapeString.trimmingCharacters(in: CharacterSet(charactersIn: "[]"))

Mochi Diffusion/Model/Scheduler.swift

@@ -9,29 +9,44 @@ import Schedulers
 
 /// Schedulers compatible with StableDiffusionPipeline
 enum Scheduler: String, CaseIterable {
-    /// Scheduler that uses a pseudo-linear multi-step (PLMS) method
-    case lcm = "LCM"
-    /// Scheduler that uses a second order DPM-Solver++ algorithm
+    case ddim = "DDIM"
+
     case dpmSolverMultistep = "DPM++ 2M"
 
     case dpmSolverMultistepKarras = "DPM++ 2M Karras"
 
+    case dpmSolverSinglestep = "DPM++ SDE"
+
     case dpmSolverSinglestepKarras = "DPM++ SDE Karras"
+
+    case eulerDiscrete = "Euler"
+
+    case eulerDiscreteKarras = "Euler Karras"
 
     case eulerAncenstralDiscrete = "Euler Ancenstral"
+
+    case lcm = "LCM"
 }
 
 func convertScheduler(_ scheduler: Scheduler) -> Schedulers {
     switch scheduler {
-    case .lcm:
-        return Schedulers.lcm
+    case .ddim:
+        return Schedulers.ddim
     case .dpmSolverMultistep:
         return Schedulers.dpmSolverMultistep
     case .dpmSolverMultistepKarras:
         return Schedulers.dpmSolverMultistepKarras
+    case .dpmSolverSinglestep:
+        return Schedulers.dpmSolverSinglestep
     case .dpmSolverSinglestepKarras:
         return Schedulers.dpmSolverSinglestepKarras
+    case .eulerDiscrete:
+        return Schedulers.eulerDiscrete
+    case .eulerDiscreteKarras:
+        return Schedulers.eulerDiscreteKarras
     case .eulerAncenstralDiscrete:
         return Schedulers.eulerAncenstralDiscrete
+    case .lcm:
+        return .lcm
     }
 }

Mochi Diffusion/Support/Extensions.swift

@@ -7,7 +7,6 @@
 
 import CompactSlider
 import CoreML
-import GuernikaKit
 import SwiftUI
 import UniformTypeIdentifiers
 

Mochi Diffusion/Support/ImageController.swift

@@ -58,6 +58,9 @@ final class ImageController: ObservableObject {
 
     @Published
     var startingImage: CGImage?
+
+    @Published
+    var maskImage: CGImage?
 
     @Published
     var numberOfImages = 1.0
@@ -257,20 +260,24 @@ final class ImageController: ObservableObject {
 
         var pipelineConfig = SampleInput(prompt: prompt)
         pipelineConfig.negativePrompt = negativePrompt
-        if let size = currentModel?.inputSize {
-            pipelineConfig.initImage = startingImage?.scaledAndCroppedTo(size: size)
+        pipelineConfig.size = CGSize(width: ImageController.shared.width, height: ImageController.shared.height)
+
+        if let size = pipelineConfig.size, startingImage != nil{
+           pipelineConfig.initImage = startingImage?.scaledAndCroppedTo(size: size)
+            pipelineConfig.inpaintMask = maskImage?.scaledAndCroppedTo(size: size)
         }
         if startingImage == nil && currentControlNets.isEmpty{
             strength = 1.0
         }
+
         pipelineConfig.strength = Float(strength)
         pipelineConfig.stepCount = Int(steps)
         pipelineConfig.seed = seed
         pipelineConfig.originalStepCount = 50
         pipelineConfig.guidanceScale = Float(guidanceScale)
         pipelineConfig.scheduler = convertScheduler(scheduler)
         for controlNet in currentControlNets {
-            if controlNet.name != nil, let size = model.inputSize, let image = controlNet.image?.scaledAndCroppedTo(size: size) {
+            if controlNet.name != nil, let size = pipelineConfig.size, let image = controlNet.image?.scaledAndCroppedTo(size: size) {
                 if ImageGenerator.shared.pipeline?.supportsControlNet == true{
                     let c = try? ControlNet(modelAt: URL(fileURLWithPath: controlNetDir + controlNet.name! + ".mlmodelc"))
                     let cinput = ConditioningInput.init(module: c!)